BackgroundCognitive impairments reduce adaptive responses and may increase the risk of falls. ObjectivesTo compare gait kinematics in older adults with cognitive impairments living in long term institutions and to identify the ability of gait kinematics to predict falls in older adults with cognitive impairments living in long term institutions. MethodsData of 23 older adults with cognitive impairments living in long term institutions were considered for this study. Fifty gait cycles were recorded during walking at a self-selected pace using footswitches sensors. The variables considered for the analysis were: speed; stride length; stance, swing and stride time; and the variability of these parameters. Fall status was recorded for a 6 month-period. FindingsMANOVA found group effect (p = 0.025) for gait kinematics comparisons. Variability of stance (p = 0.01) and swing (p = 0.012) and stride time (p < 0.001) were higher in older fallers. Speed of older fallers was 31.8% slower than those of the non-fallers (p < 0.001). The kinematic variables that were able to predict falls were: stride time variability (p < 0.001), threshold of 0.4 s, sensitivity of 50% and specificity of 100%; and gait speed (p < 0.001), threshold of 0.65 m·s−1, sensibility and specificity of 50%. InterpretationOlder adults living in an assisted living facility with a history of falls demonstrate increased kinematic variability while walking. However, the ability of gait kinematic parameters to predict falls was found to be weak. The results suggested that gait kinematic parameters are weak predictors of falls in older adults with cognitive impairments living in long term institutions.